> I was just having a quick look at the source code to FPSE and was shocked that
> the cpu emulator looks like it's actually very very compact (cpu2.cpp in the
> source, 500 lines) and just seems to have stuff like
>
> case ADDU: rd = rs rt; break;
> case SUBU: rd = rs - rt; break;
> case BNE: if (rt!=rs) { JUMP(PC immS*4); } break;
>
> Is this because MIPS is a very reduced instruction set (i.e. the decode is
> similar in each opcode)?

Maybe the overhead isn't much with a fast CPU and a reduced instruction set, but when I tried to write a Z80 emu, I chose to initialize an array of function pointers *ONCE*, and then decode opcodes this way:

opcode = readOpcodeFromMemory(PC);

(functionArray[opcode]) (parameters);

Downsides were the init time, which was kinda long, and the fact that all functions implementing opcodes shared the same signature, which was a useless parameter passing, sometimes. But no time wasted to "seek" the correct opcode implementation.

Was it a very bad idea?